对分布式的基于竞争的MAC协议中的分组冲突进行合理的分类对于有效进行冲突解析起着关键作用.文中根据冲突节点所处退避阶段的不同将冲突划分为两类:交叉冲突和同级冲突,并且通过分析和实验证明了网络中的交叉冲突是普遍存在的.传统的退避算法并未考虑这两种冲突的不同特点,而是采用同样的方式进行处理,对系统性能造成了一定影响.作者认为,对于不同的冲突类型应该予以区别对待,因此提出了冲突顺序解析算法CSR(Collision Sequential Resolution).CSR根据冲突发生的顺序,将冲突节点依次分布在一系列连续独立的基本窗口上,通过竞争窗口的离散化消除了交叉冲突;同时,通过选择合适的基本窗口大小在分组延迟和同级冲突概率之间取得折衷.仿真实验表明,同传统的退避算法相比,CSR能够在冲突次数、吞吐量、延迟以及公平性方面提供全面的性能提升.
Collision classification is crucial for collision resolution in distributed contention-based MAC protocols. Based on backoff stages the collided nodes stay, this paper classifies collisions into two categories: cross collision and intra collision, and then proves by analysis that cross collision is pervasive in networks, which is verified later by simulation results. Traditional backoff algorithms did not discriminate between these two collisions and treated them alike, which damages the system performance. However, the and thus:propose a novel backoff algorithm two collisions should be dealt with different policies featuring Collision-Sequential Resolution (CSR) to address the problem. CSR redistributes the collided nodes in a series of consecutive separated elementary Windows according to their occurrence sequence to eliminate cross collision completely. The intra collision is resolved with an appropriate elementary window size to achieve a tradeoff between delay and intra collision probability. Extensive simulations for collision times, throughput, delay and fairness demonstrate that the CSR provides comprehensive improvement compared with traditional backoff algorithms.